Viruses Regulate Microbial Community Assembly Together with Environmental Factors in Acid Mine Drainage

ABSTRACT

Viruses are widespread in various ecosystems, and they play important roles in regulating the microbial community via host-virus interactions. Recently, metagenomic studies showed that there are extremely diverse viruses in different environments from the ocean to the human gut, but the influences of viral communities on microbial communities are poorly understood, especially in extreme environments. Here, we used metagenomics to characterize microbial communities and viral communities in acid mine drainage (AMD) and evaluated how viruses shape microbial community constrained by the harsh environments. Our results showed that AMD viral communities are significantly associated with the microbial communities, and viral diversity has positive correlations with microbial diversity. Viral community explained more variations of microbial community composition than environmental factors in AMD of a polymetallic mine. Moreover, we found that viruses harboring adaptive genes regulate a relative abundance of hosts under the modulation of environmental factors, such as pH. We also observed that viral diversity has significant correlations with the global properties of microbial cooccurrence networks, such as modularity. In addition, the results of null modeling analyses revealed that viruses significantly affect microbial community phylogeny and play important roles in regulating ecological processes of community assembly, such as dispersal limitation and homogenous dispersal. Together, these results revealed that AMD viruses are critical forces driving microbial network and community assembly via host-virus interactions.

IMPORTANCE Viruses as mobile genetic elements play critical roles in the adaptive evolution of their hosts in extreme environments. However, how viruses further influence microbial community structure and assembly is still unclear. A recent metagenomic study observed diverse viruses unexplored in acid mine drainage, revealing the associations between the viral community and environmental factors. Here, we showed that viruses together with environmental factors can constrain the relative abundance of host and microbial community assembly in AMD of copper mines and polymetallic mines. Our results highlight the importance of viruses in shaping the microbial community from the individual host level to the community level.

INTRODUCTION

Viruses are one of the most diverse and abundant life forms across Earth’s ecosystem, from oceans to extreme environments (1), and they play key roles in regulating the microbial community via killing hosts (2), reprograming host metabolic pathways (3), and changing host behaviors (4). For instance, viruses suppress the bloom of fast-growing hosts (i.e., kill-the-winner dynamic) (5) and promote nutrient cycling and microbial diversity (6), which maintains the composition and structure of the microbial community. During infection processes, viruses improve the spread of adaptive genes in the microbial community through lysogenic conversion, such as antibiotic resistance genes (7), which is a critical force driving the diversification of microbial populations (8). In addition, viruses alter species interaction strength and further shape community structure by changing species abundance (9). For example, infection with phage T7 results in Escherichia coli acquiring T7 resistance, increasing the secretion of cross-fed metabolites for Salmonella enterica (10), which means that phage can indirectly influence the mutualisms in networks of cross-feeding bacteria. Furthermore, viral predation acting as a heterogeneous selection process counterbalances homogenous selection of the environment, resulting in an undominated process governing microbial community assembly, and this is supported by a recent study of fractured shale ecosystems (11). These findings reveal that viruses underpin the microbial community in multiple ways. Investigating the interactions of the viral community and the microbial community under the modulation of environments, we can better understand the outcome of ecosystems constrained by host-virus interactions, especially for extreme environments that have a relatively low-complexity microbial community.

Acid mine drainage (AMD) is characterized by low pH (pH <3) and high concentrations of metals (12), but it harbors diverse microorganisms (13), including bacteria, archaea, and viruses. During the past decades, numerous microbial species were isolated from AMD, and their genomes were well characterized, providing clues for the importance of viruses in host genomic evolution. For example, viral genome sequences were frequently detected in the genome of iron- or sulfur-oxidizing bacteria, such as Leptospirillum group II (14) and Acidithiobacillus species (15, 16). Also, heterotrophic bacteria of Alicyclobacillus species harbor various viral genes that are responsible for energy and carbohydrate metabolism in oligotrophic environments (17). Visually, cryo-transmission electron microscopy (TEM) showed that viral infection processes frequently took place in archaea cells in AMD during archaeal Richmond Mine acidophilic nano-organism (ARMAN)-Thermoplasmatales interconnectedness via cytoplasmic bridges and pili (18), implying that viruses may influence interspecies interactions. Recent advances in metagenomic research showed that 96% of viral species cannot be assigned at the genus level, and viral taxonomic richness is significantly affected by pH and microbial richness (19). These findings suggest that diverse viruses frequently interact with microbial species in AMD; however, the influences of viruses on the microbial community are poorly understood.

Using metagenomics, we characterized the microbial community and viral community in acid mine drainage derived from a copper mine and a polymetallic mine. The aims of our study addressed following three questions: (i) Does the viral community play more important roles in shaping the structure of microbial community than environmental factors? (ii) Can viruses influence topological properties of their host in the microbial interaction network and (iii) drive microbial community assembly? Our results highlight that viruses are a critical force driving microbial community assembly, and contribute more to the alpha diversity of the microbial community than environmental factors.

RESULTS

Physiochemical properties of AMD derived from copper mines and polymetallic mines.

There were significant differences in physiochemical properties of AMD between copper mines and polymetallic mines (Fig. 1; P < 0.001). For example, AMD of copper mines has a lower pH (1.3 ± 0.37) and higher temperature (38.35°C ± 3.14°C) than that of polymetallic mines with a pH of 2.45 (±0.11) and temperature of 32.58°C (±4.94°C). Meanwhile, most of the heavy metal concentration in AMD of copper mines was higher than that of polymetallic mines, such as Cu, Cr, and Cd. In addition, the concentrations of sulfate and sulfur in the AMD of copper mines were also significantly (P < 0.05) higher than those in the AMD of polymetallic mines.

FIG 1.

Boxplot of environmental factors of acid mine drainage derived from copper and polymetallic mines. NS (not significant), P ≥ 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Community structure and diversity of microbes and viruses.

For the microbial community, a total of 3,439,367 16S rRNA reads were recovered from clean metagenomic reads and clustered into 6,235 operational taxonomic units (OTUs). At phylum level, microbial communities in AMD of copper mines were dominated by Nitrospirota (77.09 ± 6.92%, Zijinshan mine) and Thermoplasmatota (93.64% ± 2.36%, Monywa mine), while those in AMD of polymetallic mines were primarily dominated by Proteobacteria (54.84 ± 28.36%). As for viruses, 2,686 viral contigs with a length ≥10 kb were identified from 4.9 million metagenomic contigs and clustered into 1,516 viral OTUs (vOTUs). The gene-sharing network inferred by vConTACT 2 showed that these viral contigs were clustered into 547 genus-level groups, none of which can be assigned with known taxonomy. Meanwhile, the accumulation curves of OTUs and vOTUs becomes stable when the number of samples increases to 10 and 5, respectively (see Fig. S1 in the supplemental material), which indicates that these data are enough to draw robust conclusions about the viral and microbial communities.

The structures of the microbial community and viral community were significantly different between copper mines and polymetallic mines (Fig. 2A; all P = 0.001). Meanwhile, viruses in AMD of copper mines have significantly higher abundance (64,400.43 ± 5,924.71) than those in polymetallic mines (41,538.60 ± 9,678.35; two-side t test, P < 0.01). However, viral community in AMD of polymetallic mines have higher species number (491.50 ± 134.34) or diversity (3.41 ± 0.71) than that of copper mines (species number: 255.17 ± 78.78; diversity: 4.36 ± 0.61; Fig. 2B; two-side t test, all P < 0.01). Furthermore, we found that viral diversity has significantly positive relationships with microbial species diversity and phylogenetic diversity (Fig. 2C; P < 0.05) in AMD of copper mines, while such a finding was not observed in polymetallic mines (Fig. 2C; P > 0.05).

FIG 2.

Structure and diversity of the microbial community and viral community. (A) Nonmetric multidimensional scaling analysis of viral and microbial communities. (B) Boxplot of the number and diversity of operational taxonomic units (OTU) and viral operational taxonomic units (vOTU). (C) Linear relationships between vOTU diversity and OTU diversity or phylogenetic diversity (PD). Green, polymetallic mine; purple, copper mine. NS, P ≥ 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001. Solid lines represent a significant linear relationship.

Environmental factors shaping the microbial community and viral community.

The microbial community and viral community were significantly associated with environmental properties in copper mines (Mantel test, all P < 0.05) but not in polymetallic mines (all P > 0.05). Specifically, in AMD of copper mines, viral communities were significantly influenced by pH, temperature, and the concentration of sulfate and most metals, such as Fe, Al, Cu, and Co (Fig. S2; all P < 0.05), and a similar pattern appeared in microbial communities (Fig. S2). However, in AMD of polymetallic mines, viral communities were mainly affected by some metals, including Fe, As, Co, and Cr (Fig. S2; all P < 0.05), while microbial communities were only significantly affected by Co and Ni (Fig. S3; all P < 0.05). Further, results of random forest regression analyses showed that, in AMD of copper mines, the key environmental factors constraining viral diversity are distinct from those constraining microbial diversity. For example, viral diversity was mostly influenced by pH and the concentration of Al and Cu (Fig. 3A), while microbial diversity was mostly affected by the concentration of S, Co, and SO₄^2– (Fig. 3B). In AMD of polymetallic mines, viral diversity was less sensitive to environmental factors than microbial diversity. For instance, microbial diversity was primarily affected by the concentration of V, Fe, and Ba (Fig. 3C), while viral diversity likely had weak associations with environmental factors (Fig. 3D; P > 0.05). Overall, results of linear regression analyses showed that viral diversity has significantly negative relationships with the concentration of Mg, Pb, and Si (Fig. S4; all P < 0.05), while microbial diversity has no such associations (Fig. S4; all P > 0.05). Moreover, results of variation partition analysis showed that, in AMD of polymetallic mines, virus community explained more variation (21.3%) in microbial community than that explained by environmental factors (8.3%; Fig. 3F), which was inconsistent with AMD of copper mines (Fig. 3E).

FIG 3.

Effects of environmental factors on microbial (OTU) and viral (vOTU) communities. (A to D) The relative importance of environmental factors in (A) OTU diversity and (B) vOTU diversity for AMD of copper mines and that in (C) OTU diversity and (D) vOTU diversity for AMD of polymetallic mines. (E and F) Variation of the microbial community explained by viruses and environmental factors in AMD of (E) copper mines and (F) polymetallic mines calculated by variation partition analysis. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Host-virus connections and abundance pattern.

In total, we found 312 host-vOTU pairs consisting of 24 metagenome-assembled genomes (MAGs) across 6 phyla and 249 vOTUs. Actinobacteria had most members infected by viruses (Fig. 4A), but hosts in the lineage of Acidobacteriota and Proteobacteria were infected by mostly diverse viruses (Fig. 4B). This result indicates that viruses can infect more clades in the lineage of Actinobacteria than others (20), but Actinobacteria hosts have lower susceptibility to diverse viruses than Acidobacteriota and Proteobacteria. Moreover, the GC content and genome size of viruses infecting bacteria of Actinobacteria, Proteobacteria, and Acidobacteriota were significantly larger than those of Archaea, such as Thermoplasmatota and Micrachaeota (Fig. 4C and D).

FIG 4.

Information of host-virus linkages. (A) Number of hosts infected by AMD viruses across various phyla. (B) Number of viral species infecting hosts across various phyla. (C) The GC content of host genomes. (D) Contig length of viruses infecting different phyla.

Further, we found that viruses regulated the relative abundance of hosts in multiple strategies. First, higher viral abundance was not accompanied by increasing relative abundance of hosts (Fig. S5), such as Actinobacteriota (1S01Z10.14) and Thermoplasmatota (1W1.18), which implies that viruses may lyse the host to maintain relative abundance of the host. Second, there was a positive linear relationship between vOTU abundance and relative abundance of their host (Fig. S5), such as Acidobacteriota (1W1.44), which implies that viruses may facilitate hosts acquiring more sources for survival, improving their adaptability and competitiveness.

Third, viruses harboring adaptive genes jointly with environmental factors regulate host relative abundance. For example, we found that one vOTU (meta_sample_2-k127_252564) processing a gene encoding vacuolar iron transporter protein (K22736; Fig. 5) has no individual effect on the relative abundance of the host (Sulfobacillus sp. C6W.96), as well as ferrous content (Table S1). However, the combined effect of viral abundance and ferrous content has a significantly positive influence on the relative abundance of the host (Table S1; P < 0.05, adjusted R² = 0.5526). Similarly, one vOTU (meta_sample_6-k127_82400) harboring a gene encoding a proton-conducting membrane transporter (Fig. 5) has no individual influence on the relative abundance of the host (Cuniculiplasma sp. 1W1.18; Table S2). After accounting for the joint effect of viruses and pH, vOTU abundance and its interaction with pH have a significantly negative influence on host relative abundance (Table S2; P < 0.05; adjusted R² = 0.1105).

FIG 5.

Gene context of two viral sequences processing auxiliary metabolic genes encoding transporters of iron and protons. Different colors represent genes involved in different functions.

Effects of viruses on the microbial interaction network.

Our results predicted that viruses have a significant influence on the microbial cooccurrence network (Fig. 6). For example, viral diversity has positive correlations with the number of network nodes and edges (Fig. 6; P < 0.05) in AMD of copper mines and polymetallic mines, but that of polymetallic mines was stronger than that of copper mines. This finding implies that the high complexity of the microbial network would be constrained by more diverse viruses. Meanwhile, viral diversity has negative correlations with average path length and network centralization (Fig. 6; P < 0.05), which indicates that diverse viruses can tighten species connections and enhance network cohesion. Viral diversity also has negative linear relationships with network modularity (Fig. 6; P < 0.05), which means that diverse viruses may disturb the modular organization of the microbial cooccurrence network.

FIG 6.

Linear relationships between viral diversity and network properties of the microbial cooccurrence network. Solid lines, P < 0.05, dashed lines, not significant.

We further showed that viral abundance has significant effects on the topological features of hosts within the cooccurrence network. For instance, viral abundance has significantly negative correlations with the betweenness of hosts affiliated with network connectors and peripherals (Fig. S6; P < 0.05) in AMD of copper mines but not with those of polymetallic mines (Fig. S6; P > 0.05). In addition, the degree of hosts as network connectors decreased with the increasing abundance of viruses, implying that viruses may weaken the connection among network modules. Also, the betweenness of hosts as peripheral nodes decreased with the raising abundance of viruses, indicating that viruses can relax the stress centrality of hosts and make the network fragile.

Effects of viruses on microbial community assembly.

Our results showed that in AMD of copper mines, the beta net relatedness index (βNRI) ranged from −4.67 to 5.43 with an average of 0.35 (±3.85; Fig. S7), indicating that microbial community assembly was dominated by stochastic processes, which was in consistent with the beta nearest taxon index (βNTI) (0.72 ± 2.29; Fig. S7). However, in AMD of polymetallic mines, βNRI ranged from 0.49 to 12.70 with an average of 5.03 (±3.23), implying that microbial community assembly was constrained by deterministic processes. Further, we found that viral community dissimilarity has a significantly positive linear relationship with βNRI in AMD of copper mines (Fig. 7A; P < 0.05), but which was negative in AMD of polymetallic mines (Fig. 7A). Consistently, environmental difference has significantly positive linear relationships with βNRI in AMD of copper and polymetallic mines (Fig. 7B), indicating that environmental selection is a critical force of deterministic processes in microbial community assembly. In addition, there were no significant linear relationships between βNTI and viral community dissimilarity (Fig. 7C) or environmental difference (Fig. 7D) in AMD of copper mines or polymetallic mines. Further, random forest regression analysis showed that the importance of viral community dissimilarity in βNRI (465.59) was similar to that of environmental difference (479.69) in AMD of copper mines. However, in AMD of polymetallic mines, the importance of viral community in βNRI (76.92) was dramatically less than that of environmental difference (176.50).

FIG 7.

The relationships between beta net relatedness index (βNRI) or beta nearest taxon index (βNTI) of microbial community and viral community dissimilarity or environmental difference. (A) There were significantly (P < 0.05) positive linear relationships between viral community dissimilarity and βNRI in AMD of copper mines, while there was a negative relationship in AMD of polymetallic mines. (B) There also were positive linear relationships between environmental difference and βNRI in AMD of copper mines and polymetallic mines. (C and D) Viral community dissimilarity (C) and environmental difference (D) have no linear relationships with βNTI in AMD of copper mines and polymetallic mines.

Furthermore, we applied a phylogenetic bin-based framework to investigate the effects of viruses on phylogenetic groups consisting of their hosts. All predicted hosts were clustered into two bins including bin21 and bin29 (Fig. S8). Bin21 contained 5 members, Actinobacteria, Firmicutes_E, Micrarchaeota, Proteobacteria, and Thermoplasmatota, while bin29 consisted of 3 members, Acidobacteriota, Actinobacteria, and Thermoplasmatota. For bin21, phylogenetic assembly in AMD of copper mines was dominated by drift (63.6%; Fig. 8A), while that in AMD of polymetallic mines was primarily constrained by dispersal limitation (64.3%; Fig. 8B). As for bin29, homogenizing dispersal (98.5%) and dispersal limitation (71.4%) dominated the phylogenetic assembly in AMD of copper mines (Fig. 8A) and polymetallic mines (Fig. 8B), respectively. Further, random forest analysis showed that in AMD of copper mines, the importance of viruses in the Raup-Crick (RC) distance of bin21 was similar to that of environmental factors (Fig. 8C). However, in AMD of polymetallic mines, the importance of viruses in the RC distance of bin21 was obviously larger than that of environmental factors (Fig. 8D). Moreover, in AMD of copper mines and polymetallic mines, the importance of viruses in βNRI or RC distance of bin29 was consistently larger than that of environmental factors (Fig. 8E and F). Furthermore, the importance of viruses and environmental factors in βNRI was larger than that of RC distance in AMD of copper mines (Fig. 8E), in contrast to that in AMD of polymetallic mines (Fig. 8F). This finding implies that viruses were constrained more on the phylogenetic structure than the taxonomic composition of bin29 in AMD of copper mines compared to that of polymetallic mines.

FIG 8.

Phylogenetic bin analyses of predicted hosts infected by AMD viruses. (A and B) Ecological processes of microbial community assembly in bin21 and bin29 in AMD of copper mines (A) and polymetallic mines (B). (C and D) The importance of viruses and environmental factors in βNRI and Raup-Crick distance (RC) of bin21 in AMD of (C) copper mines and (D) polymetallic mines calculated using the random forest regression model. (E and F) The importance of viruses and environmental factors in βNRI and RC of bin29 was calculated in AMD of (E) copper mines and (F) polymetallic mines. All random forest regression models were statistically significant (P < 0.05). In panels A and B, different colors represent different ecological processes in community assembly. Green, dispersal limitation (DL); purple, drift (DR); orange: homogenizing dispersal (HD).

DISCUSSION

A large diversity of unexplored viruses in acid mine drainage.

Exploring the distribution and genomic diversity of viruses is helpful to uncover their ecological roles and function potentials. The development of metagenomic technology gives an unprecedented opportunity to explore viral diversity across various ecosystems and largely expands the global virome catalogue. For instance, there were 2.31 million viral contigs within the biggest database of IMG/VR v3.0 in September 2020 (21), which increased by 3-fold from June2019 (22). In this database, 85.5% of them were recovered from seawater, freshwater, and host-associated environments, while merely 2.4% were from extreme environments such as hypersaline aquatic environments (21), indicating that extremophilic viruses are largely unknown (23–25). Here, we identified 2,686 viral contigs belonging to 1,516 vOTUs from 20 AMD water samples, none of which was assigned with known taxonomy through reticulate classification (vConTACT 2). Similarly, a previous study showed that 11,112 viral contigs were identified from 90 AMD sediment samples and clustered into 5,678 vOTUs, 96.0% of which cannot be assigned with taxonomy (19). These findings mean that there are largely diverse and unknown viruses in acid mine drainage.

Viruses regulate host relative abundance under the modulation of environmental factors.

Viruses play important roles in maintaining ecosystem stability by regulating the microbial community. On the one hand, viruses as predators control host abundance by lysing cells. In kill-the-winner dynamics, viruses usually kill fast-growing species and maintain host relative abundance in the community (5, 6). Our results showed that viral abundance increased but the relative abundance of their host did not change, such as heterotrophic Acidiphilium of Proteobacteria, which amount to a low proportion of the MD microbial community (13). On the other hand, viruses prefer lysogenic infection, and their density increases with host increasing density. According to the piggyback-the-winner hypothesis, viruses switch from lytic to lysogenic infection when host growth reaches eutrophic status and host population achieves high density, which can give a competitive advantage to the dominant species via viral genes and increase host relative abundance (6). We found that viral abundance has positive linear relationships with the host relative abundance of chemoautotrophic species, such as Thermoplasmatota.

Beyond these two strategies, we also found that viruses regulate host relative abundance constrained by environmental features. Viruses contribute to host competitiveness (26) and environmental adaptability (27, 28) via reprograming host metabolisms with auxiliary metabolic genes (AMGs). For example, heterotrophic bacteria of Alicyclobacillus species living in AMD enriched viral AMGs involved in energy and carbohydrate metabolism to adapt to oligotrophic conditions (17). Here, we found that the relative abundance of iron-oxidizing bacteria, such as Sulfobacillus spp., was jointly affected by ferrous ion concentration and the abundance of viruses carrying s gene encoding vacuolar iron transporter protein, which helps host to uptake and store iron and increases host competitiveness. In this case, viruses optimized host metabolic pathways and improved host growth rates; thus, viral abundance has an association with host abundance. Similarly, we also found that the relative abundance of cell-wall-deficient archaea, such as Cuniculiplasma (29), was significantly constrained by the synergism of pH and the abundance of viruses harboring a gene encoding proton-conducting membrane transporter, which was a part of proton efflux systems and improved microorganism adaptability in an extremely acidic environment (30). However, introduction of viral genes requires extra energy consumption and metabolic burden from hosts (31), in which hosts make a trade-off between competitiveness and adaptability.

Viruses are key factors in shaping the microbial cooccurrence network.

Viruses can underpin species cooccurrence patterns by influencing host behavior. In ecological networks, changes in species behaviors affect species topological roles and reshape network structure via a cascade effect (9). Our results showed that viral diversity has a significant influence on global properties of the microbial network. For example, viral diversity has significantly positive relationships with network size and average degree of nodes, which implies that diverse viruses can strengthen the interactions among microbial species. Previous studies have reported that phages not only facilitate competitive exclusions but also enhance species coexistence in the microbial community. For example, the presence of phages results in more intense competition between their host and bacterial competitors (32), which leads to greater suppression of host density. Moreover, a phage also can reinforce mutualistic interactions between its host and cross-feeding bacteria by stimulating the host to secret more cross-fed metabolites. Meanwhile, our results showed that viral diversity increases with the decrease of average shortest path length among network nodes, which means that diverse viruses can improve the efficiency of information transfer in the microbial community. These findings highlight that, apart from the direct regulation of host behaviors, viruses have a synergistic effect on non-host species to shape the microbial network.

Furthermore, diverse viruses may influence the stability of microbial network. A previous study has reported that the microbial co-occurrence network has low stability indicated by properties such as high centrality and low modularity (33). On the one hand, our results showed that more diverse viruses lowered the centrality of the network, which is a tendency to generate a network hub to a high degree, which can enhance network stability. On the other hand, high viral diversity decreased network modularity, which disturbed the module structure and weakened network stability. Similarly, a previous study showed that viral lysis pressure can decease network modularity but stimulate microbial connections within the network module. These findings imply that viruses can lower the network stability but strengthen module stability.

Roles of viruses in microbial community assembly.

Viruses play important roles in the evolutionary trajectory of microbial species, which may further influence microbial community phylogeny. Our results showed that the viral community has a significant association with the phylogenetic diversity of the microbial community, which indicates that viruses can constrain community phylogenetic assembly. Viral predation was interpreted as selection processes and counterbalances environmental selection in microbial community assembly processes (11). In acid mine drainage, the processes dominating microbial community assembly rely on the balance of the constraints of viruses and environmental factors on the microbial community. For example, in AMD of copper mines, microbial community assembly was primarily influenced by stochastic processes when viruses had similar effects on the microbial community with environmental factors. However, in AMD of polymetallic mines, microbial community assembly was dominated by deterministic processes when viruses played a more important role in shaping the microbial community than environmental factors. There are three mechanisms of how viruses encounter environmental filtering in microbial community assembly. First, viruses depress the response of host cells to environmental changes. A previous study reported that viral lysis prefers certain bacterial phylogenetic groups with a high leucine uptake rate, such as abundant taxa of Alphaproteobacteria and Betaproteobacteria in freshwater habitats (34), which means that viruses obviously constrain the response behaviors of microbial species to nutrition content. Second, viruses processing resistance genes contribute host resistance to environmental stressors, which largely weaken the environmental selection on host survival. For example, AMD Alicyclobacillus species integrate many viral genes involved in heavy metal resistance, accelerating the adaptative evolution of those species inhabiting AMD-contaminated soil (17). Similarly, at higher Cr contamination levels, lysogenic viruses harbor more auxiliary metabolic genes, improving host heavy metal detoxification in the soil microbial community (35). Third, some viruses can promote biofilm formation by providing the extracellular polysaccharides required for biofilm construction (28), which can buffer the environmental stress to microbial community. Furthermore, some lytic viruses can provide biofilm matrix (e.g., environmental DNA [eDNA]) by killing host cells and maintain the material turnover in biofilm microenvironment (28). These findings highlighted that viruses can indirectly influence community assembly and weaken environmental selection on the microbial community.

In addition, the role viruses play in microbial community assembly largely relies on host range and host sensitivity to various viruses. If viruses have a wide host range and hosts are sensitive to diverse viruses, viral predation can be interpreted as homogeneous selection that can encounter environmental filtering (11) and result in community assembly governed by drift (Fig. 8A) . In this case, viral predation also may raise dispersal limitation because immigrant populations have no specific immune systems against viruses, such as a clustered regularly interspaced short palindromic repeats along with Cas proteins (CRISPR-Cas) system and a restriction-modification (RM) system. Our results supported that assembly of phylogenetic bins was dominated by dispersal limitation when viruses played more important roles in shaping microbial community assembly than environmental factors (Fig. 8A to D). Otherwise, viruses have a narrow host range, and viral predation can be considered heterogeneous selection, which can strengthen deterministic processes in microbial community assembly. Furthermore, if viruses can improve the widespread presence of adaptive genes across the microbial community, virus lysogenization can be considered a critical force driving homogeneous dispersal. Our results showed that assembly of phylogenetic bins would be dominated by homogeneous dispersal when viruses constrain more on βNRI than environmental factors in AMD copper mines. Taken together, these findings indicate that viruses are critical forces driving phylogenetic assembly of the microbial community and can flexibly regulate ecological processes during community assembly.

Conclusions. There was a largely unknown diversity of viruses in acid mine drainage and an obvious difference in viral community structures between polymetallic and copper mines. Both viral and microbial communities were significantly constrained by environmental properties, but the primary factors shaping the viral community were significantly different from those shaping the microbial community. Furthermore, viral diversity has a significantly positive correlation with microbial diversity, and the viral community can explain more variation in the microbial community than environmental factors in AMD of polymetallic mines, implying that viruses may play more important roles in shaping the AMD microbial community than environmental factors. We also found that viruses can influence the degree of microbial species in the cooccurrence network, and viral diversity has significant associations with the global properties of the microbial cooccurrence network, such as modularity. Moreover, viruses significantly affect microbial community assembly and regulate ecological processes together with environmental selection.

MATERIALS AND METHODS

Sample collection and physicochemical analyses.

A total of 20 AMD water samples were collected from 4 mining areas, Dabaoshan polymetallic mine, Guangdong Province, China (6 samples; 113.72°N, 24.52°E), Bainiuchang polymetallic mine, Yunnan Province, China (2; 103.46°N, 23.28°E), Zijinshan copper mine, Fujian Province, China (9; 116.38°N, 25.19°E), and Monywa copper mine, Myanmar (3, 95.1°N, 22.1°E). For all samples, temperature and pH were measured, and the concentration of Fe²⁺, Fe³⁺, and SO₄^2– and the content of metal elements were determined. More details about sample description and physiochemical analyses were described by previous study (12). To thoroughly recover the putative viral contigs, we also brought 11 AMD metagenomic samples from Huaxi coal mine, Guizhou Province, China (26.52°N, 116.38°E), into the profiling of the viral community, but not for downstream statistical analyses.

DNA extraction, library preparation, and metagenome sequencing.

To research the interactions between the virus community and microbial community, we focused on those viruses adsorbing onto the cell surface or inserting into the host genome sequence, because they directly interact with host cells. For that, about 20 L of AMD liquid samples was filtered through a 0.22-μm-mesh membrane filter (Polyethersulfone (PES), Φ47 to 50 mm, 0.22 μm, PALL, USA), and then the membrane was cut into 0.2-cm² pieces for DNA extraction. DNA extraction was conducted with a PowerSoil DNA isolation kit (Qiagen, USA) according to the manufacturer’s protocol. DNA concentration was assessed with Qubit 4 (Thermo Fisher Scientific, Waltham, MA USA) and Nanodrop One devices (Thermo Fisher Scientific). DNA was dissolved in Tris hydrochloric acid buffer with pH 8.0 and frozen at −20°C.

Sequencing libraries were generated by using an ALFA-SEQ DNA library prep kit following the manufacturer’s protocol, and index codes were added. The library quality was assessed on the Qubit 4.0 fluorometer (Life Technologies, Grand Island, NY) and Qsep400 high-throughput nucleic acid protein analysis system (Houze Biological Technology Co., Hangzhou, China). The library was sequenced using Illumina HiSeq 2500 instruments (10 Gb of 150-bp paired-end reads) by Guangdong MagiGene Technology Corporation (Guangzhou, China).

Metagenome assembly and binning.

Primers of raw reads were trimmed and low-quality reads were removed by BBDuk (36). The retained high-quality reads of each sample were individually assembled using metaSPAdes (37) and MEGAHIT (–meta -k 21,33,55,77,99,127) (38). Contigs with a length of <2,000 bp were removed using Cutadapt (39), and their quality were controlled using Quast (40). For each sample, contigs were binned with MetaBAT 2 (41). The completeness and contamination of binning metagenome-assembled genomes (MAGs) was evaluated using CheckM (42). MAGs with completeness of >50% and contamination of <10% were assigned taxonomy using GTDB-Tk v1.3 software (43) against the Genome Taxonomy Database (GTDB, http://gtdb.ecogenomic.org) GTDB r95. MAGs recovered from all AMD samples were dereplicated with dRep (44) (-sa 0.95 -pa 0.9 -nc 0.30 -comp 50 -con 10).

Profiling of the microbial community.

To investigate the structure of the microbial community, including bacteria and archaea, 16S rRNA gene fragments (i.e., miTags) were extracted from metagenomic raw reads using phyloFlash v3.4.1 (45) by mapping against the SILVA small-subunit (SSU) reference database (v128) (46). Extracted 16S miTags were assembled into full-length sequences and assigned a nearest taxonomy. As a supplement of the SILVA SSU reference, nearly full-length 16S rRNA gene sequences (≥1,200 bp) were extracted from MAGs using barrnap v0.9 (47).

Full-length 16S rRNA gene sequences recovered by 16S miTags and extracted from MAGs were clustered into operational taxonomic units (OTUs) at a 97% similarity level with 90% coverage with MMseqs2 v13.451111 (48). To calculate the number of OTUs for each sample, clean metagenomic reads were mapped to the OTU representative sequences using CoverM v0.6.0 (49) in contig model (–trim-min 0.10 –trim-max 0.90 –min-read-percent-identity 0.95 –min-read-aligned-percent 0.75). To avoid sampling bias, the number of reads for each sample was rarefied to 20,539 according to the minimum sampling depth.

Profiling of the viral community.

Viral sequences were identified with VirSorter2 v2.2.3 (50) according to the SOP protocol (dx.doi.org/10.17504/protocols.io.bwm5pc86). Briefly, double-stranded DNA (dsDNA) and single-stranded DNA (ssDNA) viruses with a minimal length 10 kb were identified using VirSorter2 with a cutoff score of 0.5. Nonviral sequences or host regions were trimmed by using CheckV v0.8.1 (51). Auxiliary metabolic genes in trimmed virus sequences were identified using DRAM v1.2.0 (52). According to the SOP protocol, to avoid the false-positive results, only two categories of potential virus sequences were kept for downstream analyses: (i) viral gene > 0 or VirSorter2 score ≥ 0.95 or hallmark gene > 2 and (ii) viral gene = 0 and host gene = 0.

Viral sequences were clustered into species levels of viral operational taxonomic units (vOTUs) with 95% average nucleotide identity (ANI) and 85% alignment fraction of the shorter one, as described by a previous study (53). To calculate the abundance of vOTUs for each sample (that is, reads per kilobase per million mapped reads), metagenomic raw reads were mapped to the vOTU representative sequences using CoverM v0.6.0 in genome mode (–min-read-percent-identity 0.95 –min-read-aligned-percent 0.75 -m rpkm).

Putative host prediction.

Putative hosts of viruses were predicted using four methods based on the host reference database consisting of 444 MAGs. First, viral sequences were directly aligned with MAGs by using blastn v2.9.0+ (54) with the parameters identity, ≥70%; alignment length, ≥2,500 bp; E value, ≤10⁻³; and bitscore, ≥50. Second, viral tRNAs were identified using tRNAscan-SE v2.0.3 (55) (-A and -B) and aligned against a tRNA reference complied from 444 MAGs using blastn with the parameters identity, ≥90%; alignment length, ≥60 bp; query coverage, ≥95%; mismatches, ≤10; and E value, ≤0.001. Third, CRISPRs of MAGs were identified with CRISPRDetect v2.2 (56) and aligned against viral sequences by using blastn with the parameters identity, ≥95%; query coverage, ≥95%; mismatches, ≤1; and E value, ≤1. Finally, VirHostMatcher (VHM) (57) was used to calculate d2* values of host-virus pairs, which is background-subtracting measure based on oligonucleotide frequency dissimilarity. The MAGs were considered a potential host if d2* values were less than 0.25 and ranked among the top 10 hits of one virus as previously suggested (57).

Phylogenetic community assembly.

An unrooted phylogenetic tree of the microbial community was constructed with FastTree 2 (58) based on OTU representative sequences. The phylogenetic diversity of the microbial community was estimated using the pd function within the APE v5.6-2 package (59). To detect phylogenetic signals in environments, Mantel correlation was used to calculate the correlations between phylogenetic distance and environmental dissimilarity based on Euclidean distance, which was conducted using the cal_mantel_corr function in the microeco v0.10.1 package (60). The beta net relatedness index (βNRI) and beta nearest taxon index (βNTI) of the microbial community were calculated with the cal_ses_betampd and cal_ses_betamntd functions in the microeco package. The inference of community assembly mechanisms by phylogenetic bin null model analysis (iCAMP) (61) was further applied to access the assembly of microbial phylogenetic bins in response to virus infection.

Construction of the microbial cooccurrence network.

To investigate the effects of viruses on microbial cooccurrence patterns, we constructed a microbial interaction network via the SparCC method (62), which was carried out using the trans_network function within the microeco v0.11.0 package (60). For network reliability, OTUs with a relative abundance of less than 0.1% were removed, and the links between paired OTUs were kept when their correlation was significant (P < 0.05), and the threshold (0.63) of correlation was automatically searched via a random matrix theory (RMT)-based method, which was performed using the cal_network function. Subnetworks of each sample were extracted from the whole network according to the existing OTUs, which was conducted with the subset_network function. Nodes of the subnetwork were assigned into modules via the fast greedy algorithm a with the cluster_fast_greedy function. The cal_network_attr function was used to estimate global properties of the subnetworks, including the number of nodes and links, average degree of nodes, average path length (i.e., a mean value of the shortest path length between two nodes), average clustering coefficient (i.e., a mean value of clustering coefficient of all nodes), density, heterogeneity (i.e., asymmetry of node degree distribution), centralization, and modularity (63).

Statistical analyses.

In total, 12 and 8 ecological repetitions were used to perform statistical analyses for copper mines and polymetallic mines, respectively. The permutational multivariate analysis based on Euclidean distance was used to test the significant differences in AMD physiochemical properties between the copper mines and polymetallic mines, which was conducted using the adonis2 function in vegan v2.5-7 package (64). The significant differences in environmental factors and species diversity were tested with the Wilcoxon test. The differences in the structure of microbial communities or viral communities were analyzed by nonmetric multidimensional scaling (NMDS) (65) based on Bray-Curtis distance, which was performed using the metaMDS function. Linear relationships between microbial diversity and viral diversity were analyzed with the linear regression model, which was performed using the lm function (66). The correlations of environmental factors with the microbial community and viral community were calculated by Mantel correlation, which was carried out with the mantel_test function in the ggcor v0.9.8.1 package (67). The importance of environmental factors in OTU diversity or vOTU diversity was estimated with the random forest model, which was conducted using the randomForest function in the randomForest v4.7-1 package (68). The variation in microbial community explained by viral community and environment property was accessed by variance partitioning analysis (VPA) with redundancy analysis ordination (RDA), which was conducted with the varpart function in the vegan v2.5-7 package (64). The differences in viral features across host phyla were estimated by multiple comparisons based on least significant difference (LSD) analysis, which was carried out with the LSD.test in the agricolae v1.3-5 package (69).

Data availability.

All viral sequences, vOTUs representing sequences, and MAGs generated from the current study have been deposited in the National Omics Data Encyclopedia (NODE) database under the analysis ID OEZ008166.